R-CRMA: An Access Scheme for Ring and Bus Configurations
Original Publication Date: 1991-Jul-01
Included in the Prior Art Database: 2005-Apr-03
Publishing Venue
IBM
Related People
Nassehi, M: AUTHOR [+3]
Abstract
Disclosed is Round-based-CRMA (R-CRMA), an access scheme for LANs and MANs based on ring and bus configurations. R-CRMA is derived from CRMA [1,2,3] and provides fairness, spatial reuse, bounded access delay, and effective priority.
R-CRMA: An Access Scheme for Ring and Bus Configurations
Disclosed is
Round-based-CRMA (R-CRMA), an access scheme
for LANs and MANs based on ring and bus configurations. R-CRMA is
derived from CRMA [1,2,3] and provides fairness, spatial reuse,
bounded access delay, and effective priority.
R-CRMA can
operate on a family of configurations.
Each
configuration is defined by its topology (single ring, dual ring,
folded bus, or dual bus), transmission type (slotted or unslotted)
and multiplexing mode (buffer insertion, segmented, or neither, which
is referred to as pure). Spatial reuse is possible for
buffer-insertion or segmented multiplexing modes. The protocol is
described for a pure/unslotted/ single-ring configuration, shown in
Fig. 1, where four nodes are assumed.
The scheduler performs the
global access functions by issuing Reserve and Confirm access
commands.
The dynamics
of R-CRMA is shown in the time-space diagram of
Fig. 2. L denotes the ring latency and H
a threshold. Initially, the
scheduler issues a Reserve which accumulates requests by nodes for a
set of cycles. Upon the reception of
Reserve, the scheduler selects
a subset of these cycles, constituting a round, with a total length
not exceeding the threshold, H. It then
informs the nodes of this
subset by issuing a Confirm, followed by a Reserve for the next
round. Every node can transmit for up to
the amount of its confirmed
request. It does not forward the Reserve
until it has exhausted its
confirmed requests. It also uses the
Reserve to make new requests.
The return of Reserve to the scheduler indicates the end of the
round. The scheduler initiates the next round by issuing the next
pair of Confirm and Reserve.
For each
round, the scheduler allocates bandwidth by selecting
the subset of cycles to be confirmed.
Consider a bar representation
for node requests per round as shown in Fig. 3.
The confirmed
portion of requests are crosshatched. The area of a complete
cycle-stripe/ request-bar intersection corresponds to the maximum
transmission time each node may request per cycle. In the figure, it
is assumed that the threshold H is 8.5 times this maximum. The
scheduler confirms eac...